Machine Learning Technique for Interpretation of Infrared Spectra Measured on Polymer Modified Binders

Demand and supply of polymer modified binders (PMB) have significantly increased in the past 20 years. These binders are intended to ensure long-lasting performance of asphalt pavements. Depending on the type and magnitude of modification, PMBs can be engineer to maintain superior mechanical properties over a wide range of service temperatures. PMBs can be also used to enhance pavement surface properties such as texture, noise absorption and skid resistance, and they can be applied in the modern surface treatments for pavement maintenance. This paper utilizes one of the machine learning techniques, namely neural network, in order to classify PMBs based on their infrared (IR) signature. Such a classification is particularly justifiable when dealing with highly modified asphalt (HiMA) which is a premium PMB material with a polymer content of at least 7.5 wt%. This paper demonstrates IR measurements and interpretation of 22 different unaged asphalt binders from two producers. Based on the advanced analysis it is shown that machine learning technique can very accurately differentiate between various asphalt binder groups.